Product managers are the most interesting AI training audience I work with. They understand the technology better than almost anyone else in the building. They can explain transformer architectures at a whiteboard. And yet, when it comes to actually using AI in their daily workflows, most product teams are stuck at the same 15-20% adoption rate as everyone else. The gap between understanding AI and using AI productively is wider than most PMs want to admit.

Toni Dos Santos is Co-Founder of Spicy Advisory, where he builds AI workflow training programs for product, engineering, and go-to-market teams at mid-market and enterprise companies.

Why Product Teams Underuse AI Despite Being "Tech-Savvy"

There is a paradox at the center of AI adoption in product organizations. Product managers are typically the first to evaluate new AI tools, the first to run pilots, and the first to write internal memos about AI strategy. But Pendo's 2025 State of Product Leadership report found that only 24% of product managers use AI tools daily in their actual workflows. Amplitude's product analytics data showed a similar pattern: product teams that evaluated AI tools spent an average of 12 hours testing them but only 1.4 hours per week using them in production work.

Three dynamics explain the gap:

The "I can do it better" bias. Product managers are skilled writers and analytical thinkers. When they try an AI tool and it produces a mediocre first draft, they conclude it is faster to just do it themselves. What they miss is that a mediocre first draft you can edit in 10 minutes still beats a blank page that takes 45 minutes to fill.

The evaluation trap. PMs are trained to evaluate tools, not adopt them. They test, compare, score, and move on. The muscle for building a repeatable personal workflow around a tool is different from the muscle for assessing whether the tool is good. Many PMs have evaluated a dozen AI tools and adopted zero.

The craft identity issue. Writing PRDs, synthesizing research, and making prioritization calls are core PM skills. Using AI for these tasks can feel like outsourcing the parts of the job they take pride in. McKinsey's 2025 report on AI and the workforce found that knowledge workers in roles with strong craft identities show 2.1x higher resistance to AI-assisted workflows than those in more process-oriented roles.

User Research Synthesis at Scale

This is the workflow where AI delivers the most immediate and undeniable value for product teams. A typical PM conducts or reviews 8-15 user interviews per research cycle. Synthesizing those interviews — identifying patterns, extracting quotes, mapping insights to product themes — takes 6-10 hours of focused work. It is the task PMs procrastinate on most, which means insights get stale before they reach the roadmap.

Here is the workflow that consistently saves 60-70% of that time:

Forrester's 2025 research on product management productivity found that teams using AI for research synthesis reduced time-to-insight by 58% and reported higher confidence in their prioritization decisions because the synthesis was more comprehensive. The AI does not replace the PM's judgment about what to build. It gives them better raw material to make that judgment.

PRD and Spec Writing with AI Assistants

PRDs are where the craft identity issue hits hardest. Product managers take pride in their specs. Suggesting that AI can help write them triggers defensiveness. But the workflow is not about having AI write the PRD. It is about using AI to get from scattered notes to a structured first draft in 15 minutes instead of 90.

The effective workflow has three stages:

Stage 1 — Brain dump to structure: Dump your raw notes, meeting takeaways, and bullet points into the AI with a prompt like: "Organize the following notes into a PRD outline with these sections: Problem Statement, User Stories, Success Metrics, Scope (In/Out), Technical Considerations, and Open Questions. Do not invent information — only reorganize what I have provided, and flag any sections with insufficient input."

Stage 2 — Section expansion: Take the structured outline and expand one section at a time. For User Stories, prompt: "Based on the problem statement and notes above, draft 5-7 user stories in the format 'As a [user type], I want to [action] so that [outcome].' Use only information from my notes." For Success Metrics, prompt: "Suggest 3-4 measurable success metrics for this feature based on the stated goals. Include a target range for each."

Stage 3 — Critical review: Use the AI as a reviewer. Prompt: "Review this PRD for gaps. What questions would an engineering lead ask after reading this? What edge cases are not addressed? What assumptions am I making that should be stated explicitly?" This step alone is worth the entire workflow. Gartner's 2025 survey on product development efficiency found that PRDs reviewed by AI assistants before engineering handoff had 34% fewer clarification requests during sprint planning.

Competitive Analysis Automation

Competitive analysis is the PM task most obviously suited to AI, yet most product teams still do it manually — or worse, they don't do it at all because it takes too long. A Crayon 2025 competitive intelligence report found that 41% of product teams update their competitive analysis less than once per quarter, and 18% have no formal competitive tracking process.

The AI-assisted competitive workflow:

This workflow takes competitive analysis from a sporadic, time-intensive project to a lightweight weekly habit. Pragmatic Institute's 2025 product management survey found that teams with regular competitive analysis cadences made roadmap decisions 27% faster because they spent less time debating market context in planning meetings.

Roadmap Prioritization Frameworks with AI

Prioritization is the PM task where AI is most useful and least intuitive. The value is not in having AI make the prioritization decision — that remains a human judgment call. The value is in having AI structure the inputs so the decision is better informed.

Here is a workflow that integrates AI into a RICE or weighted scoring framework:

Step 1 — Data assembly: Feed in your backlog items with whatever context exists: customer requests, support ticket volumes, revenue impact estimates, engineering effort estimates. Prompt: "For each of the following backlog items, summarize the available data on Reach, Impact, Confidence, and Effort. Flag any items where data is insufficient for confident scoring."

Step 2 — Scenario modeling: Prompt: "Given these scored items, show me three different roadmap scenarios: one optimized for customer retention, one optimized for new revenue acquisition, and one optimized for technical debt reduction. For each scenario, show the top 5 items and the trade-offs of what gets deprioritized."

Step 3 — Assumption stress-testing: Prompt: "Challenge the assumptions in Scenario 1. What would change if the revenue impact estimate for [Item X] is 50% lower than projected? What if engineering effort for [Item Y] doubles?" This creates a sensitivity analysis that most product teams never run because it is too time-consuming manually.

"Product managers don't need AI to think for them. They need AI to organize the chaos so they can think more clearly. The best PM workflows I've seen use AI as a structuring layer, not a decision layer." — Toni Dos Santos, Co-Founder, Spicy Advisory

Sprint Planning and Retrospective Summaries

These are the two ceremony-adjacent tasks where AI delivers the highest ratio of time saved to effort invested. Neither is glamorous. Both consume hours that product managers would rather spend on strategic work.

Sprint planning prep: Before planning, feed in the backlog, the sprint goal, and the team's velocity data. Prompt: "Based on a velocity of [X] story points and a sprint goal of [goal], recommend a sprint backlog from the following items. Flag any items that have unresolved dependencies or missing acceptance criteria." This turns a 30-minute prep task into a 5-minute review task.

Retrospective synthesis: After the retro, feed in the team's notes (or the retro board export from Miro, FigJam, or EasyRetro). Prompt: "Synthesize these retrospective notes into: top 3 things that went well, top 3 improvement areas, and 2-3 specific action items with suggested owners. Identify any recurring themes from the past 3 retros if previous notes are provided." Atlassian's 2025 State of Teams report found that teams running AI-assisted retrospective synthesis were 2.4x more likely to follow through on action items because the outputs were clearer and more specific.

The common thread across all these workflows: AI does not replace product judgment. It eliminates the low-value preparation work that sits between the PM and the high-value decision. Every hour saved on synthesis, formatting, and first-draft generation is an hour redirected to customer conversations, strategic thinking, and cross-functional alignment — the work that actually differentiates great product teams.

Ready to train your product team on AI workflows that stick? Spicy Advisory runs hands-on AI training sessions built specifically for product managers — covering research synthesis, PRD writing, competitive analysis, and prioritization. Every participant leaves with working workflows, not slide decks. Explore our product team AI training program.

Frequently Asked Questions

Which AI tool is best for product management workflows?

There is no single best tool. ChatGPT and Claude are strong for PRD writing and research synthesis. Microsoft Copilot integrates well if your team lives in the Microsoft ecosystem. The tool matters less than the workflow — a well-structured prompt in any major AI assistant will outperform a poorly structured prompt in the "best" tool.

Will AI replace product managers?

No. AI automates the preparation and structuring work that PMs spend 40-60% of their time on — synthesis, first drafts, data organization. The core PM skills of customer judgment, strategic prioritization, and cross-functional influence are not automatable. AI makes PMs more effective at the work that matters most.

How do I get my product team to actually adopt AI workflows?

Start with user research synthesis — it delivers undeniable time savings with minimal craft identity friction. Once PMs experience saving 5+ hours on a research cycle, they become open to trying AI in other workflows. Avoid starting with PRD writing, which triggers the most resistance. Build momentum with quick wins first.

How long does it take for AI workflows to become habitual for product teams?

Expect 3-4 weeks with structured reinforcement. Product teams adopt faster than average because they understand the technology, but they need the embedding cadence — daily practice, async feedback, and office hours — to move from experimentation to habit. Without reinforcement, most PMs revert to manual workflows within 2 weeks.